Proposition 2.1 c), p. 5 — lower epi-convergence: g(x) ≤ liminf e_{λk}g(xk) when xk ⇀ x, λk ↘ 0
ProvedMoreauChance.Consistency.proposition_2_1_cconvex-analysismoreau-envelopep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1variational-convergence
Let be a separable real Hilbert space and convex and lower semicontinuous. Let weakly in , and let be a nonincreasing sequence of positive numbers with (that is, ). Then
This is the "liminf" half of the epi-convergence of Moreau envelopes, in the weak topology. In the paper it is the step that passes from the values of the regularized problems back to the original objective.
Formalization Note The liminf is taken in the extended reals.
Preamble
import Mathlib import Definitions.Def_MoreauChance_Consistency_Basic open Filter Topology open scoped InnerProductSpace
Formal statement
namespace MoreauChance.Consistency
theorem proposition_2_1_c
{H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H] [CompleteSpace H]
[TopologicalSpace.SeparableSpace H]
(g : H → ℝ) (hg_convex : ConvexOn ℝ Set.univ g) (hg_lsc : LowerSemicontinuous g)
(xk : ℕ → H) (x : H) (lamk : ℕ → ℝ) (hxk : WeakConv xk x) (hlam_pos : ∀ k, 0 < lamk k)
(hlam_anti : Antitone lamk) (hlam_lim : Tendsto lamk atTop (𝓝 0)) :
(g x : EReal) ≤
Filter.liminf (fun k => moreauEnv (fun z => (g z : EReal)) (lamk k) (xk k)) atTop := by sorry
end MoreauChance.Consistency
Source
van Ackooij, Pérez-Aros, Soto & Vilches, Inner Moreau envelope of nonsmooth conic chance constrained optimization problems, arXiv:2301.09803v1, p. 5, Proposition 2.1 c)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.